A Semi-supervised Sensing Rate Learning based CMAB scheme to combat COVID-19 by trustful data collection in the crowd

Area of Science:

• Computer Science
• Artificial Intelligence
• Cybersecurity

Background:

• Mobile crowd sensing (MCS) platforms face challenges in recruiting trustworthy workers.
• Existing methods often assume worker quality is known or easily determined.
• Strategic workers may submit fake data (false data attacks) to maximize revenue, compromising data authenticity.

Purpose of the Study:

• To propose an incentive mechanism for recruiting multiple unknown and strategic workers in MCS.
• To address the challenge of evaluating data authenticity from potentially dishonest workers.
• To develop a system that ensures truthfulness and individual rationality in worker recruitment.

Main Methods:

• Modeling worker recruitment as a multi-armed bandit reverse auction problem.
• Developing a UCB-based algorithm to balance exploration and exploitation of worker Sensing Rates (SRs).
• Introducing a Semi-supervised Sensing Rate Learning (SSRL) approach with distinct supervision and self-supervision phases.
• Integrating SSRL with the bandit reverse auction mechanism (SCMABA) for efficient SR acquisition.

Main Results:

• The proposed SCMABA mechanism theoretically guarantees truthfulness and individual rationality.
• SCMABA demonstrates outstanding performance in simulations using real-world data traces.
• The method effectively separates exploration and exploitation phases for worker recruitment.

Conclusions:

• SCMABA offers a robust solution for recruiting reliable workers in MCS environments.
• The incentive mechanism successfully mitigates issues related to strategic worker dishonesty and false data attacks.
• The study highlights the effectiveness of combining semi-supervised learning with bandit algorithms for MCS worker management.